- Wild elephant intrusion detection, alerting sensor system determines elephant presence within 35 m
The design and implementation of a novel wild elephant intrusion detection and alerting sensor system which uses the odour of elephant urine and its’ ear-flap sound to determine the presence of an elephant in the vicinity of 35 metres has been presented and proposed to mitigate the human-elephant conflict (HEC) existing in South and South East Asian countries including Sri Lanka.
This "'eleAlert' - a sensor for detecting elephant intrusions at boundary villages" was designed by Engineer M.A.B.S. Weerawardhana, Eng. Dr. D.N. Balasuriya (Senior Lecturer at the Department of Electrical and Computer Engineering of the Faculty of Engineering Technology of the Open University) and Dr. T.S.P. Fernando (Senior Lecturer at the Department of Zoology of the Faculty of Natural Sciences of the same University) and explained in the Fourth Issue of the 55th Volume of the Engineer: Journal of the Institution of Engineers, Sri Lanka in December, last year (2022).
R. Sukumar's "The Asian elephant: Ecology and management" mentions that the Asian elephant (Elephas maximus), is one of the three living species of the largest land animal on earth. The Asian elephant is also an endangered species. Its’ habitats range from South Asia to South East Asia. The Asian elephant is tightly linked to the history, culture, religion and mythology of many countries in South and South East Asia for over 2,500 years. Tamed Asian elephants have, in the past, been used in transportation and agriculture. However, the majority of the Asian elephant population is still limited to the wild.
Recent human activities such as deforestation, excessive farming, large scale land and infrastructure development, and the segmentation of elephant habitats, have paved the way for the downfall of this pachyderm. P. Fernando's "Elephants in Sri Lanka: Past, present and future" and P. Fernando, M.A. Kumar, C. Williams, E. Wikramanayake, T. Aziz and S. Singh's "Review of the HEC mitigation measures practised in South Asia" mentions that the elephant population in Asia has been reduced from around 100,000 to 40,000 over the duration of a few decades. Furthermore, limiting elephant habitats, and habitat segmentation in Asian countries, have not only fuelled the demise of the elephant population, but in recent years, the harmony between humans and elephants has also been hindered. In Sri Lanka, large scale infrastructure projects such as the Mahaweli accelerated project in the 1980s and the highway network expansion projects during the last two decades have worsened the conflict between humans and elephants.
As a result of the loss of habitats and food, elephants have constantly intruded on human populated areas, destroying not only the farmlands and properties, but also human lives. Every year, hundreds of human lives are lost due to elephant attacks. On the other hand, shooting and the use of poison to protect farmlands from elephant attacks is very common in Asia. Thus, many elephants are killed by humans every year.
Nevertheless, it is our duty to protect this giant for the future. Every possible measure should be taken to protect the Asian elephant and for this, there is a requirement to minimise the HEC. Thus, many Asian countries have enlisted “decreasing the HEC” as one of their Millennium Development Goals.
There are a variety of techniques used in Asia for minimising the HEC by restricting elephant access to human occupied areas. However, since the elephants too have lost their habitats, restriction methods have not been very fruitful. On the other hand, the elephant being a very intelligent mammal, finds ways of breaching the protections and intruding into human populated areas. Consequently, B.M.O.A. Perera's "The HEC: A review of the current status and mitigation methods" notes that recent attention has been focused on alerting elephant intrusions in advance in order to avoid potential conflicts. With a timely alert, people can either reach safety or take necessary action to drive the elephants away, without harming them. This in turn saves the lives of both humans and elephants.
Hence, Weerawardhana et al. developed an elephant intrusion alert sensor.
There are many elephant intrusion restricting mechanisms used in the world today. They can be broadly classified into two groups. In one group, the techniques are to directly restrict the elephant intrusion. The electric fence is one of the most popular techniques used around the world to restrict elephant intrusions into human populated areas. A metal wired fence is supplied with an electrical voltage high enough to produce a shock to the elephant. The electric shock is not fatal to the elephant, but it is a psychological barrier to cross. However, electric fences are quite expensive to implement and operate, having a capital cost ranging around Rs. 300,000 per kilometre. On top of this capital cost, it incurs an operational cost for the electricity used in energising the fence. Thus, in many developing countries, the usage of electric fences is limited to only the most critical areas. On top of all these, it has been reported that elephants place tree branches on the electric fence to break the fence.
Elephants have an inherent fear of bees and of the scent of chillie. According to a Gabonese study and S. Hedges and D. Gunaryadi's "Reducing the HEC: Do chillies help deter elephants from entering crop fields?" in parallel to the electrical fences, beehives and chillie pot fences are also employed in African countries to restrict the elephants from intruding into villages. In these fences, either the beehives or chillie powder pots are hung, spaced by several metres. Though this technique provides promising results against the African elephants, the technique has not proven to be very efficient against the Asian elephant. On the other hand, the maintenance of beehives and the chillie pots are very difficult tasks under the tropical climates of South and South East Asia. The elephant intrusion restricting systems may also cause a limitation on elephant movement which can in turn erupt as an intrusion at some other place.
Apart from these elephant restricting systems, there are several elephant intrusion detection systems employed in Asia and Africa. They use either the images or videos captured (K.S.P. Premarathna, R.M.K.T. Rathnayaka and J. Charles's "Image detection system for elephant directions along with the forest border areas" and M. Zeppelzauer's "Automated detection of elephants in wildlife video") or the infrasound calls generated by elephants ("Eloc: Locating wild elephants using low cost infrasonic detectors" by A. Sayakkara, E.N. Jayasuriya, T. Ranathunga, C. Suduwella, N. Vithanage, C. Keppitiyagama, K. De Zoysa, K. Hewage and T. Voigt, and M. Zeppelzauer and A.S. Stoeger's "Establishing the fundamentals for an elephant early warning and monitoring system”) to detect the presence of an elephant. Nevertheless, the vision based systems need to have a good visibility of the area without foliage cover which cannot be guaranteed in tropical Asian countries. On the other hand, elephants may not always emit infrasound calls, hence using infrasound calls as the detection mechanism is not always successful.
Thus, Weerawardhana et al. propose an elephant detection system based on multiple parameters, namely, the odour of elephant urine and the ear-flap sound made by elephants. Furthermore, with the use of a multi parameter intrusion alert system and a support vector classification to classify the sensor reading to “elephant present” and “no elephant or elephant absent” categories, Weerawardhana et al. expect a much more accurate intrusion detection than a single sensor based detection.
The proposed elephant intrusion detection and alert system named as “eleAlert” is meant to be employed at the boundary of human populated areas and it consists of several important subunits, namely, an elephant urine odour sensor and an ear-flap sound sensor, both of which feed into the main control unit or classifier, which in turn emits an alarm.
On average, an elephant urinates more than 15 times a day and the amount of urine passed at a time may range even up to nine litres, according to E. Wiedner, A.R. Alleman and R. Isaza's "Urinalysis in Asian Elephants". It is also observed that during urinating, a considerable amount of urine is stuck on the elephant’s body parts. Hence, even when the elephant moves to a different location, the elephant emits a considerable odour of urine.
The main chemical substance in the urine of any mammal is ammonia which is very high in elephants compared to other mammals. In order to verify the response of an Ammonia sensor to animal urine, several tests were conducted at the Dehiwala and Pinnawala Zoological Gardens using an integrated ammonia sensor module. The results of the tests to measure the strength of the urine odour at different distances from the animal were shown for six different animals which are commonly found around boundary villages, namely, the elephant, the wild buffalo, the deer, the leopard, the domestic cow and the elk. It was clear that the odour of urine is very much higher in elephants than in all the other animals, having a considerable odour even at the maximum tested distance of 35 metres.
Having a body mass ranging up to five tons, the elephant’s body generates a very high amount of heat. A. Narasimhan's "Why do elephants have big ear-flaps?" mentions that in the absence of sweat glands, the primary mechanism of the elephant’s excess heat dissipation is through the ears. Elephant ears are equipped with a large surface area and an extensive vascular network to facilitate this process. Thus, the elephant periodically flaps its ears, which generate a periodic sound signal, and it can be heard even at a considerable distance away.
As the second parameter measurement, Weerawardhana et al. developed an ear-flap sound capturing circuit which consists of three principal components, namely, the microphone (of a special module based kind), the low noise amplifier (an ultra low noise operational amplifier) and the band-pass filter (a two stage, cascaded filter which has a band-pass filter response in the said frequency range), all of which have a good response in the frequency band of interest and a very high load resistance. The low noise amplifier takes an input signal from the microphone in the range of microvolts to millivolts and amplifies the same to a volts based range without much added noise. The filter is the most critical component in the ear-flap sound detector as it allows the demarcation of the ear-flap signal from other noises.
Though the ear-flap sound is a possible parameter to detect the presence of an elephant, in the Dry Zone outdoor environments where most of the elephant intrusions occur, the other surrounding noises such as the sound of the wind is at a relatively high level. Therefore, an experiment was conducted with the presence of an elephant to determine variation with the distance from the elephant. Furthermore, a test at the same premises was conducted in the absence of an elephant. The results of the aforementioned tests clearly depict a direct current output in the range of volts in the presence of an elephant. The signal can be clearly detected over noise, up to a distance of approximately 10 metres. Since the sensor nodes are placed at the boundary of the perimeter to be protected, this range is adequate in practical applications.
The main control unit of the “eleAlert” system considers the inputs from the urine odour sensor and the ear-flap sound detector and makes an intelligent decision whether an elephant is present or not. If an elephant is present, the control system energises an alarm siren.
The training data set is limited in this particular application; hence, a support vector machine (SVM) supervised learning was used to determine a boundary between the “present” and “not present” ranges. Per R.O. Duda, P.E. Hart and D.G. Stork's "Pattern classification", the use of the SVM classifier is expected to produce better results with the limited training data set, than any other classifier. The SVM is a supervised learning technique for classifying a set of test data into two groups.
The urine sensor and the ear-flap sound detector were implemented to measure the two parameters in the presence and absence of an elephant. Then, the pairs of measurements were taken for the known cases, where an elephant was present within a 35 metres distance in 31 samples and the elephant was not present in another 20 samples. Test results under tested conditions show that the overall elephant detection system produces an accuracy of 90% in the classification. Misclassifications are all false positive cases, and are mainly contributed by the erroneous noisy readings given by the ear-flap sound sensor.
The use of a joint classification based on multiple parameters has improved the accuracy. It is also verified that this accuracy is superior to those of the single parameter based classifications.
Though the alternative elephant detection systems available provide higher accuracies in image and video based systems, they obtain this high accuracy only when an image or video of the elephant is available, while the proposed detection system does not require the visibility of the elephant. At the same time, the infrasound calls based elephant detector achieves the tabled accuracy only up to a 10 m distance. Hence, the proposed detector is much superior over the existing alternative elephant detection systems. The proposed system will be of value to mitigate the HEC existing in many parts of South and South East Asian countries.
"Other possible further improvements include interconnecting all geographically spaced ‘eleAlert’ nodes to form an internet of things based wide-area elephant intrusion monitoring system with which the relevant authorities can remotely monitor a large potential threat area with ease. Moreover, combining vision and the proposed techniques can be expected to have an improved accuracy in elephant detection," Weerawardhana et al. posited.